Technical Summary: Making Sense of Mixed Signals: The Impact of (In)consistent Performance Feedback on Corporate Digital Technology Innovation

1. Problem Statement and Research Gap

In the era of digital transformation, Digital Technology Innovation (DTI) is a critical source of competitive advantage. However, the process is characterized by high uncertainty and path dependence, creating managerial ambiguity regarding whether digital initiatives are succeeding or require strategic redirection. While prior research has identified antecedents of DTI (e.g., technological capabilities, resource endowments), it largely treats DTI as a homogeneous process. This oversight neglects a dual structure comprising mitigation innovation (focused on efficiency, stability, and preventive reinforcement) and adaptation innovation (focused on flexibility, strategic adjustment, and reactive renewal).

Furthermore, existing literature often assumes firms pursue single performance goals. In reality, organizations pursue multiple goals that generate inconsistent performance feedback (e.g., primary performance below aspiration while secondary performance exceeds it). This creates a "fog of feedback" characterized by interpretive ambiguity. The specific cognitive mechanisms through which firms interpret these mixed signals and how executive attention influences the resulting innovation strategies remain underexplored. This study addresses these gaps by examining how consistent and inconsistent performance feedback influences the heterogeneity of DTI responses and how CEO attentional vigilance moderates these effects.

2. Theoretical Framework

The study integrates the Behavioral Theory of the Firm (BTOF) and the Attention-Based View (ABV).

• BTOF: Organizations are boundedly rational and adjust search behaviors based on performance relative to aspiration levels. The study posits that multi-goal pursuit creates conflicting signals (a "fog of feedback") that shape whether firms engage in problem-driven search (mitigation) or opportunity-driven search (adaptation).
• ABV: Executive attention is a pivotal cognitive resource. The study introduces CEO Attentional Vigilance, defined by two dimensions:
  • Attention Quantity: The breadth of attention allocated to innovation cues.
  • Attention Stability: The persistence and coherence of focus over time.
    The framework suggests that vigilance acts as a cognitive stabilizer, determining how ambiguous feedback is translated into strategic innovation directions.

3. Methodology

Data and Sample

• Source: Chinese A-share listed companies from 2007 to 2023.
• Screening: Excluded Special Treatment (ST) or Particular Transfer (PT) firms, financial industry companies, and firms with missing data.
• Data Sources: China Research Data Service Platform (CNRDS) for innovation data; China Stock Market and Accounting Research (CSMAR) for financial and governance data.

Variable Measurement

• Dependent Variables (DTI): The study distinguishes between Mitigation DTI (preventive, stability-oriented) and Adaptation DTI (reactive, resilience-oriented).
  • Measurement: A hybrid approach combining factor analysis of quantitative indicators (R&D, patents) with a BERT-based machine learning semantic classification. A pretrained Chinese BERT model, fine-tuned on manually labeled corporate disclosures, classifies patent abstracts into mitigation or adaptation categories based on domain-specific lexicons (e.g., "predictive maintenance" vs. "disaster recovery").
• Independent Variables (Performance Feedback):
  • Below-Aspiration Feedback (FBB): Primary performance (e.g., ROA) falls below a weighted aspiration level (historical + industry median).
  • Inconsistent Feedback (FBI): Primary performance is below aspiration, but secondary performance (e.g., R&D intensity) is above aspiration.
  • Consistent Negative Feedback (FBC): Both primary and secondary goals fall below aspiration.
• Moderating Variables (CEO Attentional Vigilance):
  • Attention Quantity (EA): Ratio of radical innovation keywords to total innovation keywords in CEO communications (speeches, annual reports).
  • Attention Stability (VA): The inverse of the standard deviation of attention quantity across four quarters, measuring focus consistency.
• Control Variables: Firm size, age, board characteristics, ownership concentration, leverage, ROE, asset turnover, inventory intensity, institutional ownership, and Tobin's Q.

Empirical Model

The study employs fixed-effects panel regression models to examine the impact of feedback types on mitigation and adaptation DTI, controlling for firm, industry, and year effects. Interaction terms are included to test the moderating roles of CEO attention quantity and stability.

4. Key Results

Main Effects

• Below-Aspiration Feedback: When primary performance falls below aspirations, firms significantly increase mitigation DTI (preventive reinforcement) but decrease adaptation DTI (strategic renewal). This suggests a behavioral shift toward stabilizing existing systems under pressure.
• Inconsistent Feedback: When primary performance is low but secondary performance is high, the pattern reverses. Firms decrease mitigation DTI and increase adaptation DTI. The positive secondary signal provides psychological safety and resource slack, encouraging adaptive renewal despite primary shortfalls.
• Consistent Negative Feedback: Sustained underperformance on both goals reinforces mitigation innovation while constraining adaptation.

Moderating Effects of CEO Attentional Vigilance

• Attention Quantity: High CEO attention quantity weakens the positive relationship between below-aspiration feedback and mitigation DTI (by diffusing focus) and strengthens the negative relationship between below-aspiration feedback and adaptation DTI (by reducing cognitive depth for experimentation).
• Attention Stability: High CEO attention stability weakens the negative relationship between inconsistent feedback and mitigation DTI (by maintaining focus on the primary shortfall) and strengthens the positive relationship between inconsistent feedback and adaptation DTI (by enabling the integration of conflicting signals into a coherent renewal strategy).

5. Significance and Contributions

Theoretical Contributions

1. Advancing Behavioral Theory: The study moves beyond single-goal assumptions to explain how organizations navigate the "fog of feedback" arising from multi-goal pursuit. It reinterprets feedback ambiguity not as noise, but as a distinct cognitive condition that shapes the direction of search (mitigation vs. adaptation).
2. Refining the Attention-Based View: By distinguishing between attention quantity and stability, the study demonstrates that attention is a dynamic mechanism. It reveals that excessive breadth (quantity) can be detrimental under pressure, while stability is crucial for resolving ambiguity and sustaining adaptive strategies.
3. Methodological Innovation: The paper bridges computational methods and behavioral strategy by using BERT-based semantic analysis to operationalize the heterogeneity of innovation. This allows for a granular, behaviorally grounded distinction between mitigation and adaptation that traditional patent counts cannot capture.

Practical Implications

• For Managers: Performance feedback should be viewed as interpretive rather than purely evaluative. Structured routines (e.g., cross-functional reviews) can help balance preventive reinforcement with adaptive reconfiguration.
• For CEOs: Attention is a strategic resource. In dynamic environments, CEOs must structure their attention to avoid diffusion. Periodic focus reviews and strategic reflection can maintain cognitive coherence when facing mixed signals.
• For Governance: Firms should develop evaluation systems that link digital innovation to both preventive and adaptive objectives, reducing overreaction to temporary fluctuations and encouraging experimentation.

Limitations

The authors acknowledge that the study focuses on Chinese A-share firms, which may limit generalizability due to institutional and cultural specificities. Additionally, the cross-sectional nature of the panel data does not fully capture the dynamic evolution of feedback-innovation relationships across different stages of digital transformation. Future research is suggested to explore longitudinal designs and comparative studies across different institutional contexts.